agent-vault
A HTTP credential proxy and vault for AI agents like Claude Code, OpenClaw, Hermes, custom agents + harnesses, and more.
What is agent-vault?
A HTTP credential proxy and vault for AI agents like Claude Code, OpenClaw, Hermes, custom agents + harnesses, and more.
Overview of agent-vault
What it is
Agent Vault is an open-source HTTP credential proxy and vault developed by Infisical. It acts as a credential broker that sits between AI agents and the external APIs they call. Instead of giving agents direct access to credentials, Agent Vault stores them and attaches credentials to outbound HTTP requests as they pass through the proxy. This design aims to eliminate the risk of credential exfiltration caused by prompt injection attacks on AI agents. The project is no longer recommended for new setups; users are directed to use Infisical Agent Vault instead.
Key features
- Credential Brokering: Substitutes dummy header values (e.g.,
__anthropic_api_key__) with real credentials or replaces auth headers on outbound requests, so agents never hold real credentials. - Pluggable Credential Stores: Supports backing the vault with an external secrets store like Infisical, enabling features such as dynamic secrets.
- Transparent Integration: Bootstraps agents' environment to use
HTTPS_PROXYand routes all underlying requests through Agent Vault's MITM architecture, working with existing tools like MCP, CLI, SDK, and API. - Purpose-Built Design: Built specifically for credential brokering with AI agents, including a dedicated CLI, multi-tenancy, and an agent-specific roadmap.
- Egress Filtering: Controls which agents can access which services and API endpoints, since authenticated requests flow through Agent Vault.
- Request Logging: Inspects authenticated traffic to monitor and diagnose agent behavior.By default, requests not matching any service forward as plain proxy traffic; a vault can be set to strict deny mode (
unmatched_host_policy=deny) to reject them with a 403 error.
Use cases
Agent Vault works with various AI agent use cases, including secure remote coding agents (e.g., Claude Code configured to proxy through Agent Vault, with credentials like ANTHROPIC_API_KEY and GITHUB_PAT stored in the vault), secure all-purpose agents (e.g., OpenClaw, Hermes), secure custom agents built with custom harnesses, and secure ephemeral sandboxes where an orchestrator mints a temporary token for a sandboxed agent to use.
Getting started
Install Agent Vault on macOS (Intel + Apple Silicon) or Linux (x86_64 + ARM64) using the script:bashcurl --proto '=https' --proto-redir '=https' --tlsv1.2 -fsSL https://get.agent-vault.dev | shStart the server with a master password:bashexport AGENT_VAULT_MASTER_PASSWORD=your-passwordagent-vault server -dOr deploy with Docker:bashdocker run -it -p 14321:14321 -p 14322:14322 -e AGENT_VAULT_MASTER_PASSWORD=your-password -v agent-vault-data:/data infisical/agent-vaultThe server starts the HTTP API on port 14321 and a transparent HTTP/HTTPS proxy on port 14322.
License and pricing
The license is not stated in the provided metadata or README. The project is listed as open source with no pricing information.
Based on the project's public README and repository metadata.